Some content of this application is unavailable at the moment.
If this situation persist, please contact us atFeedback&Contact
1. (WO2003085842) METHOD FOR ITERATIVE HARD-INPUT FORWARD ERROR CORRECTION
Note: Text based on automatic Optical Character Recognition processes. Please use the PDF version for legal matters

METHOD FOR ITERATIVE HARD-INPUT FORWARD ERROR CORRECTION

BACKGROUND OF THE INVENTION

Field of the Invention

[0001] The invention relates to the field of communications. More specifically, the invention relates to error correction in communications.

Background of the Invention

[0002] hi a traditional communications networks, forward error correction (FEC) is

used to protect transmitted information in signal processing. In the case of algebraic block

codes (e.g. the Reed Solomon code), redundancy symbols are added to the information

symbols before transmission of a block of symbols. A network element that receives the transmission can correct transmission errors as longs as the number of corrupted symbols does not exceed a certain threshold given by the special construction of the code.

[0003] Iterative algorithms have been developed for decoding in systems where a soft input containing analogue information representing the reliability of a received symbol is

available.
[0004] However, soft input decoding techniques generally do not apply in optical

networks where a hard decoder input is given by a bit or byte stream. Moreover, a soft

input-decoding algorithms have relatively poor performance for lower probability of

transmission errors, as is the case in optical networks.

BRIEF DESCRIPTION OF THE DRAWINGS
[0005] The invention may best be understood by referring to the following description and accompanying drawings that are used to illustrate embodiments of the invention. In

the drawings:

l [0006] Figure 1 is a diagram of exemplary network elements according to one embodiment of the invention.

[0007] Figure 2 is a an exemplary flowchart for encoding data according to one embodiment of the invention.

[0008] Figure 3 A is a diagram illustrating exemplary interleaving of two code classes in a matrix according to one embodiment of the invention.

[0009] Figure 3B is a diagram illustrating exemplary interleaving of a third code class into the matrix of Figure 3 A according to one embodiment of the invention.

[0010] Figure 3C is a diagram illustrating an alternative exemplary interleaving of two code classes according to one embodiment of the invention.

[0011] Figure 4 is a flow chart for the decoding of data according to one embodiment of the invention.

[0012] Figure 5 is a diagram of components of a line card of a network element for according to one embodiment of the invention.

[0013] Figure 6A is a diagram illustrating an exemplary embodiment of the iterative decoder 505A of Figure 5 according to one embodiment of the invention.

[0014] Figure 6B is a diagram illustrating an exemplary embodiment of the iterative decoder/dewrapper and encoder/wrapper 509 according to one embodiment of the invention.

[0015] Figure 7A is a diagram illustrating an exemplary embodiment of an iterative decoder with backward annotation according to one embodiment of the invention.

[0016] Figure 7B is a diagram illustrating an exemplary embodiment of an iterative decoder with forward annotation according to one embodiment of the invention.

DETAILED DESCRIPTION OF THE DRAWINGS
[0017] In the following description, numerous specific details are set forth to provide a thorough understanding of the invention. However, it is understood that the invention may be practiced without these specific details. In other instances, well-known circuits, structures and techniques have not been shown in detail in order not to obscure the invention.
[0018] Figure 1 is a diagram of exemplary network elements according to one embodiment of the invention, h Figure 1, a network element 101 is coupled with a network element 103. A wrapping/encoding unit 107, which will later be described in more detail herein, in the network element 101 receives data 105. The data includes a set of information symbols. One or more bits may represent each of the information symbols. The wrapping/encoding unit 107 adds overhead space to the data 105 (referred to as wrapping). The wrapping/encoding unit 107 initially fills part of the overhead space with administrative information. The administrative information may include patterns for code synchronization. The wrapping/encoding unit 107 then fills the overhead space in the data 105 with redundancy symbols of interleaved code classes, which is later described in more detail herein. The wrapping/encoding unit 107 passes the wrapped data 106 to an optical transmitter 109. hi alternative embodiments, a single physical unit or separate physical units may perform the wrapping and encoding. The optical transmitter 109 converts the wrapped data 106 from an electrical signal to an optical signal 108. The optical transmitter 109 then transmits the optical signal 108 from the network element 101 to the network element 103.

[0019] An optical receiver 111 in the network element 103 receives the optical signal 108. The optical receiver 111 converts the received optical wrapped data 106 into an electrical signal. The decoding unit 113 iteratively decodes the wrapped data 106. The decoding unit 113 passes the decoded wrapped data 108 to a decoding/de- wrapping unit 115. The decoding unit 113 and the decoding/de- wrapping unit 115 will later be described in more detail herein. The decoding/de- wrapping unit 115 further decodes the wrapped data 106 received from the decoding unit 113. The decoding/de- wrapping unit 115 also removes the administrative information and overhead space from the wrapped data 106. The decoding/de-wrapping unit 115 then outputs the data 105. A single physical unit or individual physical units may perform the decoding and de-wrapping.
[0020] Iteratively decoding optically transmitted data with interleaved code classes reduces the bit error rate in the data that is output after decoding and de- wrapping. Chart 1 shown below illustrates input bit error rate (BER) versus output BER with three exemplary encoding schemes. The input BER is the probability of a single bit transmission error whereas the output BER is the probability of a single bit being corrupted after decoding a data. The dotted line labeled "RS (255, 239)" illustrates the performance of a Reed-Solomon encoding scheme. The dashed line labeled "simple RS" illustrated the performance of a simple Reed-Solomon based block product encoding scheme. The solid line represents the performance of an encoding scheme implementing iterative hard input decoding of two interleaved code classes with two iterations. As illustrated in chart 1, the encoding scheme that implements iterative hard input decoding of interleaved code classes outperforms the other exemplary encoding schemes at input BERs that typically occur in optical transmissions.

"RS (255,239)"
"Simple Product decoding"
invented Algorithm"


[0021] Figure 2 is a an exemplary flowchart for encoding data according to one embodiment of the invention. At block 201, the wrapping/encoding unit 107 determines an encoding scheme to be the current encoding scheme. An encoding scheme a forward error correction (FEC) encoding algorithm (e.g., BCH encoding, Reed-Solomon encoding, etc.) and the parameters used with the encoding algorithm. In one embodiment, the set of parameters and procedures may be predefined. In another embodiment, the
wrapping/encoding unit 107 may retrieve the parameters and encoding algorithm from memory. In an alternative embodiment, a user may select the parameters and the encoding algorithm. In another embodiment, the wrapping/encoding unit 107 may select parameters and/or the encoding procedure from different parameters and/or encoding procedures stored on the wrapping/encoding unit 107 and/or a separate storage unit.

[0022] At block 203, the wrapping/encoding unit 107 receives data. At block 205, the wrapping/encoding unit 107 processes the data with the current encoding procedure and the current parameters to generate a set of code words (i.e., the data along with
redundancy data) of a first code class. The term code class refers to all code words generated with a certain encoding procedure and a certain set of parameters. Different code classes may be generated by the same encoding algorithm, but different parameters. Different code classes may also be generated by the same parameters, but different encoding algorithms. Different code classes may also be generated by different encoding algorithms and different sets of parameters.
[0023] At block 211, the wrapping/encoding unit 107 determines a next encoding scheme to be the current encoding scheme. An alternative embodiment may select another set of parameters and another encoding algorithm. Another embodiment may use the currently selected parameters and select another encoding algorithm.
[0024] At block 213, the wrapping/encoding unit 107 generates the next code class with the current encoding scheme. At block 215, the wrapping/encoding unit 107 determines if the encoding is complete. If the encoding is not complete, then control flows to block 211. If the encoding is complete, then the wrapping/encoding unit 107 transmits the encoded data at block 217. The symbols of the data are encoded so that each symbol is a member of at least one code word of each code class.
[0025] While the flow diagrams in the Figures show a particular order of operations performed by certain embodiments of the invention, it should be understood that such order is exemplary (e.g., alternative embodiments may perform certain of the operations in a different order, combine certain of the operations, perform certain of the operations in parallel, etc.). For example, block 201 and/or block 211 may not be performed if the set of parameters are predetermined. In addition, block 203 may be performed before block 201. In another embodiment, block 201 and block 211 are performed sequentially or in parallel.

[0026] Encoding of symbols so that each symbol is a member of at least one code word of each code class my be illustrated with a matrix. Figures 3 A - 3C are diagrams illustrating interleaving of code classes.
[0027] Figure 3 A is a diagram illustrating exemplary interleaving of two code classes in a matrix according to one embodiment of the invention. In Figure 3 A, wrapped data 301 is processed by an encoding procedure 305. The encoding procedure 305 generates a first code class illustrated as a set of π2 code words arranged as rows in a matrix 309. The code words of the first code class have a block length of m and k\ symbols. The redundancy symbols for the first code class are illustrated in the matrix 309 as row redundancy symbols 311 (n\ - k\ redundancy symbols for each row). The first code class corresponds to the first dimension of the matrix 309.
[0028] A second dimension of the matrix 309 is then processed by an encoding procedure 306. As previously stated, alternative embodiments may process the matrix 309 with another encoding procedure, with another encoding procedure and another set of parameters, etc. The encoding procedure 306 generates a matrix 315. The columns of the matrix 315 are the code words of the second code class. The code words of the second code class have a block length of « , kι information symbols, and n2 - k% redundancy symbols for each code word, which are illustrated as column redundancy symbols 313. The second code class includes
code words (i.e.,
columns in the matrix 315). The second code class includes code words comprised of the row redundancy symbols 311 and redundancy symbols for correction of the row redundancy symbols 311.
[0029] Figure 3B is a diagram illustrating exemplary interleaving of a third code class into the matrix of Figure 3 A according to one embodiment of the invention. In Figure 3B, an encoding procedure 316 generates a matrix 319. The encoding procedure 319 encodes a third dimension of the matrix 306 to generate the matrix 319 with a third code class. The matrix 319 includes third dimension redundancy symbols 317. The third dimension redundancy symbols 317 correspond to each diagonal of the code words of the third code class.
[0030] Figure 3C is a diagram illustrating an alternative exemplary interleaving of two code classes according to one embodiment of the invention. Figure 3C illustrates the encoding procedure 105 generating a stream of interleaved code classes as a two-dimensional field 321. Field 321 includes row code words and column code words.
Unlike the matrix 315 of Figure 3 A, the field 321 includes multiple code words in each row. In the field 321, the code words of the first code class do not align against a single code word of the second code class.
[0031] Interleaving multiple code classes provides for improved forward error correction. Although FEC encoding schemes provide for correction of a limited and known number of transmission errors, interleaving multiple code classes enables iterative decoding to overcome this limitation of FEC encoding schemes.
[0032] Figure 4 is a flow chart for the decoding of data according to one embodiment of the invention. At block 401 a network element receives data. At block 402, the network element decodes the last code class. At block 403, the network element decodes the next code class (i.e., the code class encoded prior to the last code class). At block 405, the network element determines if all of the code classes have been decoded. If all of the code classes have not been decoded, then control flows to block 403. If all of the code classes have been decoded, then at block 407 the network element determines if the syndrome for each code class is equal to zero. If the syndrome is not equal to zero for each code class, then control flows from block 407 to block 402. If the syndrome for each code class is equal to zero, then at block 409 the network element removes the redundancy symbols from the received data.
[0033] Figure 4 illustrates how interleaving multiple code classes enables iterative decoding to overcome the limitations of current FEC encoding schemes. For example, assume the encoding procedures 305 and 306 individually provide for correction of t\ and t2 transmission errors, respectively. If the number of errors in a code word of the first code class exceed t\, then that code word typically could not be corrected. Since multiple code classes are interleaved and then iteratively decoded, corrected errors in the second code class possibly enable corrections in the first code class in a subsequent iteration. In other words, correction of an incorrect symbol that is at an intersection of code words of the first and second code class may reduce the number of errors in code word of the first code class below ti.
[0034] As previously indicated, the order of operations illustrated in Figure 4 is exemplary. For example, block 405 may not be performed because the number of code classes is known. Alternative embodiments may perform block 407 differently. A flag may be set as soon as a syndrome is calculated that does not equal zero. A total sum for all syndromes may be maintained, h addition, another embodiment may accept a certain level of error in the data, hi another embodiment, blocks 402 and 403 may be performed in parallel.
[0035] Although the environment described in Figure 4 continues iteratively decoding until all errors in the data have been corrected, embodiments of iterative hard input forward error correction may iterate through the code classes a predefined number of times.
[0036] Figure 5 is a diagram of components of a line card of a network element for according to one embodiment of the invention. In Figure 5, a line card 500 includes an optical receiver 501 receives wrapped data as an optical signal. The optical receiver 501 converts the optical signal into an electrical signal. The optical receiver 501 then passes the wrapped data in electrical signal form to a deserializer 503. The deserializer 503 arranges the wrapped data for iterative decoding. The deserializer 503 then passes the wrapped data to be processed by a series of iterative decoders 505 A - 505F. Each of the iterative decoders 505 A - 505F perform at least one iteration of decoding all code classes of the wrapped data. The iterative decoder 505F passes the wrapped data to an iterative decoder/dewrapper and encoder/wrapper 509. The iterative decoder/dewrapper and encoder/wrapper 509 performs at least one more iteration of decoding all of the code classes of the data and dewraps the data. The iterative decoder/dewrapper and
encoder/wrapper 509 then outputs the data.
[0037] The iterative decoder/dewrapper and encoder/wrapper 509 also receives data to be transmitted. The iterative decoder/dewrapper and encoder/wrapper 509 wraps the received data as described in Figure 1 and encodes the received data as previously described in Figures 1-2. The iterative decoder/dewrapper and encoder/wrapper 509 then passes the wrapped and encoded data to a serializer 511. The serializer 511 arranges the wrapped data for transmission. The serializer 511 then passes the serialized wrapped data to an optical transmitter 513. The optical transmitter 513 converts the serialized wrapped data from an electrical signal to an optical signal and transmits the optical signal.
[0038] The line card 500 and/or the components of the line card 500 include one or more machine-readable media. A machine-readable medium includes any mechanism that provides (i.e., stores and/or transmits) information in a form readable by a machine (e.g., a computer). For example, a machine-readable medium includes read only memory (ROM); random access memory (RAM); magnetic disk storage media; optical storage media; flash memory devices; electrical, optical, acoustical or other form of propagated signals (e.g., carrier waves, infrared signals, digital signals, etc.); etc.
[0039] The embodiment illustrated in Figure 5 is scalable by the number of iterative decoders coupled together. In addition, the illustrated embodiment provides substantial board space savings by implementing the encoder and the decoder as a single component. Various embodiments may implement the iterative decoders differently.
[0040] Figure 6 A is a diagram illustrating an exemplary embodiment of the iterative decoder 505 A of Figure 5 according to one embodiment of the invention. In Figure 6A, the iterative decoder 505A is for decoding two code classes, hi Figure 6A, a column decoder 601 of the iterative decoder 505A receives wrapped data. The column decoder 601 decodes each column of the received data. After decoding each column of the received data, the column decoder 601 passes the data to a data reorderer 603 A. The data reorderer 603 A rearranges the data received from the column decoder 601 from a column perspective to a row perspective. The data reorderer 603 A then passes the reordered data to a row decoder 605. The row decoder 605 decodes the received data as rows of a matrix. The row decoder 605 then passes the data to a data reorderer 603B. The data reorderer 603B rearranges the data from the perspective of rows to columns. The data reorderer 603B then passes the reordered data to the next iterative decoder.
[0041] Figure 6B is a diagram illustrating an exemplary embodiment of the iterative decoder/dewrapper and encoder/wrapper 509 according to one embodiment of the invention. In Figure 6B, the iterative decoder 505F passes data to the iterative
decoder/dewrapper and encoder/wrapper ("hybrid unit") 509. The data passes through the column decoder 601 A, data reorderer 603 A, and a row decoder 605 A, similar to each of the iterative decoders 505A-505F. The row decoder 605A passes the data to the dewrapper 611. The dewrapper 611 dewraps the data similar to the dewrapping previously described in Figure 1. Unit 509 also includes a wrapper 613. The wrapper 613 receives data to be transmitted and wraps the data similar to the wrapping previously described with respect to Figure 1. The wrapper 613 passes the wrapped data to a row encoder 615. The row encoder 615 fills in some of the space added by the wrapper 613 with row redundancy symbols. The row encoder 615 then passes this data to a data reorderer 603B. The data reorderer 603B reorders the data from a row perspective to a column perspective. The data reorderer 603B then passes the reordered data to a column encoder 607. The column encoder 607 fills the rest of the space added by the wrapper 613 with column redundancy symbols. The column encoder 607 then passes the data out of the iterative decoder/dewrapper and encoder/wrapper 509.
[0042] Figure 7A is a diagram illustrating an exemplary embodiment of an iterative decoder with backward annotation according to one embodiment of the invention. The iterative decoder illustrated in Figure 7A is an exemplary three code class iterative decoder. The iterative decoder illustrated in Figure 7A includes syndrome computation units 703A - 703C. Alternative embodiments may include more or less syndrome computation units. Error pattern computation units 709A - 709C are coupled with the syndrome computation units 703A - 703C. The iterative decoder 505A receives data having three code classes. The received data is stored in a FIFO 701 A. The received data is also sent to the syndrome computation units 703A - 703C.
[0043] The syndrome computation unit 703A computes the syndrome for the third code class. The syndrome computation unit 703A passes the syndrome to an error pattern computation unit 709 A. The error pattern computation unit 709A calculates error patterns for error correction and also calculates a back annotation. The corrections determined by the error pattern computation unit 709A are applied to the data stored in the FIFO 701 A. The resulting data is stored in a FIFO 701B. The calculated back annotation is passed from the error pattern computation unit 709A to the syndrome computation units 703B and 703 C. Since syndrome computation is linear, the syndrome computation units 703 A -703 C may compute their syndrome in parallel, hence the syndrome computation unit 703B has calculated a syndrome for the second code class. The syndrome computation unit 703B adds the received back annotation to its computed syndrome. The syndrome computation unit 703B then passes the syndrome with back annotation to the error pattern computation unit 709B. The error pattern computation unit 709B performs the same task as the error pattern computation unit 709A for the second code class.
[0044] The error pattern computation unit 709B applies the computed error correction information to the data stored in the FIFO 701B. The resulting data is stored in a FIFO 70 lC. The error pattern computation 709B passes the back annotation computed for the second code class to the syndrome computation unit 703C. The syndrome computation unit 703 C should have computed a syndrome for the first code class and added the first code classes back annotation. The syndrome computation unit 703 C then adds the back annotation for the second code class to its syndrome and passes the computed syndrome to the error pattern computation unit 709C. The error pattern computation unit 709C determines error correction information and applies the information to the data stored in the FIFO 70 lC. The resulting data is then passed to the next iterative decoder.
[0045] Iterative decoding of interleaved code classes with backward annotation reduces latency from decoding each code class. The error patterns of a code class Ci may be computed immediately after error pattern computation of a code class C has completed.
[0046] Figure 7B is a diagram illustrating an exemplary embodiment of an iterative decoder with forward annotation according to one embodiment of the invention. The iterative decoder 505A illustrated in Figure 7B is similar to the iterative decoder 505A illustrated in Figure 7 A, except Figure 7B illustrates a buffer 711. The buffer 711 temporarily hosts computed annotations to be transmitted to the next iterative decoder. Alternative embodiments may store the computed annotation in a plurality of buffers, may pass the computed annotations directed to the next iterative decoder, etc. Forward annotation further reduces latency.
[0047] As previously described, iterative hard input decoding of interleaved code classes enables forward error correction for optical transmissions.
[0048] While the invention has been described in terms of several embodiments, those skilled in the art will recognize that the invention is not limited to the embodiments described. The method and apparatus of the invention may be practiced with modification and alteration within the spirit and scope of the appended claims. The description is thus to be regarded as illustrative instead of limiting on the invention.